\begin{table}[h]
\centering
\caption{Baseline ownership of ICS does not predict purchase of intervention ICS \label{app:tab_baseline_improved_predict_purchase}}
\begin{adjustbox}{max width=\textwidth}
\begin{threeparttable}
\begin{tabular}{lcc}
\toprule
& (1) & (2) \\ \cmidrule(lr){2-3}
& \multicolumn{2}{c}{$\mathbbm{1} \left( \text{Purchased intervention ICS} \right)$} \\ \midrule
\({TREATMENT}_j\)&        0.51***&        0.45***\\
            &     (0.037)   &     (0.058)   \\ \addlinespace
\(\mathbbm{1}\left( \text{Owned ICS at baseline} \right)\)&       0.018   &      0.0081   \\
            &     (0.012)   &    (0.0083)   \\ \addlinespace
\({TREATMENT}_j \times \mathbbm{1}\left( \text{Owned ICS at baseline} \right)\)&       0.031   &      0.0089   \\
            &     (0.056)   &     (0.089)   \\ \addlinespace
\({NGO}_j\) &               &      -0.030   \\
            &               &     (0.027)   \\ \addlinespace
\({TREATMENT}_j \times {NGO}_j\)&               &        0.12*  \\
            &               &     (0.071)   \\ \addlinespace
\({NGO}_j \times \mathbbm{1}\left( \text{Owned ICS at baseline} \right)\)&               &     -0.0074   \\
            &               &     (0.012)   \\ \addlinespace
\({TREATMENT}_j \times {NGO}_j \times \mathbbm{1}\left( \text{Owned ICS at baseline} \right)\)&               &       0.042   \\
            &               &      (0.11)   \\ \addlinespace
\midrule
Control mean&           0   &           0   \\
\(N\)       &         943   &         943   \\
Adjusted \(R^2\)&        0.24   &        0.24   \\
District fixed-effects & Yes & Yes \\
\bottomrule
\end{tabular}
\begin{tablenotes}
{\setlength\labelsep{0pt}
\footnotesize
\item \textit{Notes}. The outcome variable is an indicator that equals 1 if household \(i\) in hamlet \(j\) purchased at least one of the two ICS promoted during the intervention. Column (1) presents aggregated results; results are disaggregated by NGO and non-NGO villages in column (2). Standard errors (in parentheses) are clustered at the hamlet level. \sym{*} \(p<0.10\), \sym{**} \(p<0.05\), \sym{***} \(p<0.01\).}
\end{tablenotes}
\end{threeparttable}
\end{adjustbox}
\end{table}
